ReplyEssay
A Different Economic Theory of Infrastructure
Lawrence LessigÜ
Simple ideas fix public policy. Simple ideas, taken for granted, by a generation that rules. These ideas were learned. They did not come naturally. They were taught on the basis of the best that was known, at the time this generation last learned. They are not argued for. They are not disputed. They set the background against which public policy decisions get made.
There is a set of simple ideas that now guides telecommunications policy. At its core is a view about the utility of regulation. Regulation, this view holds, is disfavored. More precisely, a very good reason is needed if private ordering is to be disturbed. Thus, markets should be left alone unless some strong reason for intervention is shown. Market failure alone is not sufficient since government failure can defeat any gain that government intervention might seek.
I agree with these simple ideas. But I also believe that with respect to networks, there is a gap in our understanding about when regulation makes sense. There is not yet a good theory for explaining this gap, nor will there be one until economists frame such a theory in their own language. For now, there is only a set of powerful intuitions, but powerful intuitions do not compete with simple ideas.
Brett Frischmann takes this debate beyond powerful intuitions. In An Economic Theory of Infrastructure and Commons Management, Frischmann offers a model for understanding the infrastructure of telecommunications networks.[1] His model teaches a different communications policy than that schooled by simple ideas. It shows us the value in regulating these networks differently.
This is an important lesson to learn. The United States
continues to fall behind other major industrial nations in broadband
penetration. Four years ago we stood at the lead. Today,
we are thirteenth.[2]
Part of the difference between us and these other nations is our attitude about
the governmentís place in securing broadband penetration. We have left the matter
solely to private actors whereas other nations have supplemented private action
with government support. Our policy fits the reigning simple ideas about how
such markets work. Frischmannís article shows us the possible error in these simple
ways.
My aim in this Eessay
is to summarize and extend. Frishmannís article is rich and complex. Frischmann
clears the way for important new work in economics. I point to some additional
work left to be done.
I. NONRIVALROUS RESOURCES AND
INFRASTRUCTURE
Resources are different, and rivalrousness is one dimension of difference. For some resources, their consumption rivals the opportunity for others to consume the same resource. For other resources, it does not. If you drink my coffee, I cannot. Coffee is rival. If I recite your poem, that does not interfere with your opportunity to recite the same poem. Poems are not rival.
One simple idea that has captured a generation of policy
makers is that nonrivalry creates trouble.[3]
The trouble is not static. Instead, it affects the resourceís supply. Nonrival
(NR) resources invite free -riding.[4]
And if the resource is the sort that needs effort to create—if it is
songs rather than sunshine—then free -riding
can sap the incentives needed to supply the NR resource. If you can free -ride
off of what I create, then my incentive to create is less than it would be if
you paid for the benefit you received. Because of free- riding,
I create less. Free -riding
thus costs something, its name notwithstanding. It costs the incentive that
would otherwise exist had there been payment for the benefit, and that cost
reduces its supply.
The proper response to a cost is to find a way to eliminate
it. The standard response is to eliminate free- riding
by granting an exclusive right. If access to the resource can be made conditional—if
excludability can been engineered into the
frame—then access to the resource can be conditioned on paying for the
benefit. The marginal incentive to produce can be preserved if exclusive rights
can be secured.
This is a perfectly fine solution for some resources. In particular,
it is a fine solution for ěconsumption goods.î If
the NR resource is simply meant to be consumed, then there is a benefit to
solving the free -rider
problem (to align incentives), and
not much cost. There is, however, some cost—the cost of whatever
mechanism is invoked to solve the free -riding
problem. So long as the benefit to incentives outweighs that cost, it makes
sense to restrict the NR resource to those who are willing to pay.
So much is commonplace among the simple ideas that I referred
to at the start. But it is here that Frischmann begins drawing his map. He
helps to distinguish cases where the ordinary solution (stop free- riding)
works, and cases where it might not. Some resources are meant to be consumed (ěconsumption
goodsî). Other resources are inputs into
the production or creation of other goods; these resources are ěintermediate,î
and they complicate the question of whether the ordinary solution to the free- riding
problem is really a helpful solution. For if these intermediate goods are used
to create other goods, then blocking access to them through an exclusive right
could block the creation of these other goods. The cost of solving the free- riding
problem is therefore not just the cost of the mechanism that solves that
problem. Rather, the cost now includes the loss produced by not being able to
create these other goods.
There is a simple response, however, to this potential loss. If the NR resource is an input to some other good, then for the same reason that NR-consumption resources can be regulated by the market, NR-input resources could be regulated by the market as well. The consumer of that second good could simply pay the price necessary to assure access to the input. The problem is the same as the NR-consumption good, one step removed.
For example, imagine that an ad agency wants a jingle for a
new campaign. It invites composers to submit songs. As with any creative work,
such songs would be NR resources: —once
composed, they could be consumed by others without diminishing the amount
available to anyone else. If nothing ěprotectedî
my composition, then you could copy it, and undercut me in selling it to the
agency.[5]
To solve that problem, the law gives me an exclusive right (a copyright) once I
compose my song. You then can no longer (legally) free- ride
on my work.
On paper, however, my song is not really worth anything to the ad agency. Its value is made possible when someone records it. The composition itself therefore is a consumption good. The composition is an input (in the notation I have developed here, a NR-input). Thus, the exclusive right the law gives me might be thought to create a barrier for the musicians who would record it. They now cannot get access (legally) unless they pay me.
But that is not a barrier within a market economy. It is a
means for allocating the NR-input to its highest valued users. My good may be
an input, but those who need the input can pay me for it. It is a NR resource,
and even an intermediate good, but the property law does not interfere with any
creativity. Again, so long as the gain in incentives to me outweighs the cost
of the restrictions, both primary and secondary, solving the free- rider
problem makes sense.
Among NR-inputs, there is a further distinction to draw. A
subset of all NR-inputs has one further feature: they are generic rather than
specific inputs to the creation of other goods. Such NR-generic-inputs feed
many different ultimate goods. These ultimate goods can be both commercial and
noncommercial. It is these NR-generic-inputs that Frischmann suggests we call
ěinfrastructure.î[6]
ěInfrastructureî is
thus that set of intermediate goods for which three conditions are true:— the
resource is (1) NR; (2) an input into some other good; and (3) an input into ěa
wide range of goods and services, including private goods, public goods and
non-market goods.î[7]
It is the third condition that creates a potential problem
for the standard solution to the free- riding
problem. Remember that the solution to the second condition
(2) was premised on a well functioning
market.: gGiving
a property right to my NR-input was no barrier so long as a market could
properly regulate access by other goods. The third conditionCondition
(3) creates the possibility that this market will not function
well. If the NR-input is sufficiently generic, and if it is an input into a
sufficiently diverse range of goods, then the market will not regulate access
to the good well. An exclusive right in these cases may do more harm than good.
The opportunity cost of restricting access, in other words, may outweigh any
gain.
There is an even further distinction to be drawn here. If an infrastructure is a NR-generic-input, then infrastructures can be of three different kinds. There could be infrastructures that feed exclusively commercial goods, and resources that feed public and social goods in addition to commercial goods. If the NR-generic-input feeds public and social goods, then our reasons for being sanguine about the restrictive effect of an exclusive right disappear. Again, the exclusive right is (relatively) costless if and only if there is a market to regulate it. But NR-generic-inputs that supply social and public goods are infrastructures that cannot depend on a market to regulate them. The nature of these ultimate goods means markets will either be absent or incomplete. Hence, the opportunity cost of an exclusive right may be greater than its benefit.
Again, if the infrastructure exclusively supports commercial goods, then the market may adequately regulate its supply. Antitrust concerns may infect that regulation, and antitrust law must therefore continue to police such markets. But if there is no antitrust concern, then there is no reason not to leave this resource to the market. It will, as well as any mechanism would, reckon the value that would produce the necessary incentives.
We are thus at the end of the string of careful distinctions
that sets up the argument that Frischmannís article delivers. fFigure
1 maps these distinctions.
Each diamond marks a distinction. Resources are either rival
or NR. If they are NR, they are either consumption or intermediate. If they are
intermediate, they are either specific, or generic.
If they are generic, they support either commercial goods exclusively, or
public or social goods as well. The column on the left are those categories for
which the market is presumptively an adequate mechanism.[8]
The row at the bottom are those goods for whichwhere
even well functioning markets would fail adequately
to adequately value the resource.
Finally, the box at the bottom includes all goods properly considered ěinfrastructure.î
Figure 1: Mapping Infrastructure
Careful analysis thus identifies the category of infrastructure that motivates Frischmannís normative argument. That argument, in turn, comes in two parts.
II. INFRASTRUCUTRE, OPEN
ACCESS, AND MARKET FAILURE
I have compressed rich detail into a very simple map. The map
is no substitute for the detail. The map, however, helps speed identification
of the types of infrastructure that presumptively would suffer market failure.
Call this that category
ětype C infrastructure.î That category sets the stage for Frischmannís central,
normative point: type C infrastructure should be supplied through an ěopen
accessî regime.[9]
The first part of that claim—that type C infrastructure
suffers market failure—will be uncontroversial enough if we think about
traditional examples of this infrastructure. Roads, telephone networks, and
governance systems are all NR-generic-inputs that feed both commercial and
noncommercial ends. To the extent that they serve noncommercial ends, the
ability of the market to adequately reckon infrastructure supply is drawn into
doubt. T—that
is the market failure. With each of these traditional resources, there is a
long history of the government playing a central role in assuring the resource
supply.
Frischmannís map shows how that tradition generalizes: more infrastructure than traditional infrastructure will present the same market failure. If highways are infrastructure, then we can see why the information superhighway is infrastructure. If information highways are type C infrastructure, then we can see why they too would present market failure.
This is not the end of Frischmannís argument, but even here, important conclusions about current policy debates emerge. Frischmann points to one such debate in a note at the very end of his article.[10] This is the story of wireless broadband in Philadelphia.
The city of Philadelphia announced a program to fund the
deployment of a wireless Internet infrastructure within the city.[11]
More than fortyAlmost sixty
percent of Philadelphia neighborhoods had no access to broadband Internet at
all.[12]
Thus, this collective response seemed fairly sensible to anyone believing that
Internet access is an important social and economic good.
Yet before the project could even begin, the state legislature
passed a law, promoted heavily by private network providers, that banned
municipalities from providing wireless broadband connectivity.[13]
That law responded to a simple idea: if the market could supply a good, that
was sufficient to conclude
that the government need not, and therefore,
should not provide that good.
Frischmannís analysis shows us why that conclusion does not follow. The market will undoubtedly provide some broadband access. If Internet access, however, is a form of type C infrastructure—as it plainly is—then the access that the market provides will be less than what is socially optimal. Just as private roads would be fewer than public, or UPS would be more restrictive that the U.S. Postal Service, Internet provided exclusively by a private market is likely to be Internet that is socially inadequate.
III. EVALUATING OPEN ACCESS AND
GOVERNMENT REGULATION
So far there is little in Frischmannís account that I would criticize, and there is even less with which I disagree. It is Frischmannís final proposition, however, that is the most important to his argument, and this final proposition is not yet proven.
Type C infrastructure, Frischmann demonstrates, presents a
market failure. That failure invites us to interrogate other mechanisms by
which type C infrastructure might be regulated. Frischmann describes two.
One alternatives: is government
regulation; the other is and ěopen
access.î Either would value type C infrastructure
differently from than the
market. In principle, both could better optimize access to type C
infrastructure. Thus, what we need to carry Frischmannís
argument to its final point, we need is a
way to know which alternative—open access or government—is better.
Now, of course, the simple idea that I described at the start is enough to bias this question in Frischmannís favor. But Frischmann cannot rely on that simple idea. He needs something more to map the conditions under which one solution is superior to the other. Yet the argument so far offers little more than a sketch.
Consider a particular recommendation that occupies Frischmann
discusses: the regulation of access to the Internet. The Internet
is plainly type C infrastructure. It follows from the argument summarized in
Part II that we cannot rely on the market alone to determine access to such a
resource. The market will undervalue the social value of such a network. More importantly,
the market would skew the development of the network because it could not see
such alternative value. An alternative regulator is required, and the question
is which regulator is proper.
Frischmann picks open access as the regulator, by which he means
a regime that assures that the network will remain open for any application or
any content to compete equally on this network. Such competition is the consequence
of the Internetís ěend-to-endî design.
Frischmann argues—rightly, it seems to me—that we should preserve
this end-to-end design.
But we have not yet seen the argument that proves that intuition.
Frischmann points to two alternatives that, from his perspective, would be
inferior.[14]
Neither the market alone, nor the government, could do the work that end-to-end
performs. The government could not do it because the government is bad at
picking winners. The market should not do it, at least where last- generationís
winners could veto their own competition. These market and government failures
lead Frischmann to recommend something else: open access.
Again, I agree with the conclusion, but we will need more to convince those who do not. There is no doubt a cost to either the government or the market in making this choice. But we need a clearer sense of the parameters for deciding when open access is a solution. Moreover, we need a better way to measure the benefit. Put differently, what is the sensitivity of this particular conclusion? On what does it depend? What would make it different?[15]
IV. END-TO-END, ěINNOVATION,î
AND INFRASTRUCTURE
Finally, let me consider the one part of Frischmannís argument
where he means to disagree with what I have written. I am not convinced thatthere
we have a meaningful disagreement.
Frischmannís argument ends with the type C infrastructure that currently presents the most hotly contested policy battles, the Internet. As Frischmann describes, the Internet was born with an end-to-end architecture.[16] That architecture shifts intelligence in this network to the ends, or edge, so far as that is possible, and seeks to keep the network itself as simple as possible. The original motivation for that design was adaptability. A large, decentralized network would grow most easily if network owners did not need to approve every service designed for this network. End-to-end thus became the moniker by which this commitment to network simplicity is described.
In fact, however, as Barbara van Schewick demonstrates, there have been at least two end-to-end arguments in the history of the network.[17] Both arguments describe where functionality should be placed in the network. The first and more technical determines placement on the basis of necessity.[18] This argument maintains that functionality that could only be performed completely at the edge of the network should not be placed within its core.[19] That requirement later became more of a policy: there was a general preference that functionality be shifted to the edge, even if in principle, it could live in the core.
Understood as policy rather than rule, the end-to-end principle is a thumb on the scale of any network design. All things being equal, the policy says, select an end-to-end design over one that is not. In some versions, the policy is more constitutional: like the First Amendment imposes a strong preference against speech regulation, end-to-end imposes a strong preference against designs that interfere with it.
Frischmann strongly favors this end-to-end principle, as do I. He rightly links this technical design to an infrastructure commons that it produces. Tim Wu and I reach a similar conclusion, but for different reasons: our focus is the innovation that such a design encourages, and policies to defend end-to-end, in turn, are defended on the basis of this innovation.[20]
Frischmann, however, argues that ěinnovationî is too narrow.[21] It connotes to him commercial innovation only. And, of course, as commercial infrastructure is the weakest ground on which to argue that there is market failure, it would weaken the argument for end-to-end to see it solely in commercial terms. Something broader is at stake, Frischmann maintains. Not just innovation, but infrastructure.
But I do not see ěinnovationî
as tied to commerce at all. Indeed, in the sense I use the term, the most
important Internet ěinnovationsî have
very little to do with commerce. Blogs, wikis, and even Internet Relay ChatIRC,
are all ěinnovationsî
enabled by the end-to-end architecture of the Internet.
My aim is to preserve the opportunities for these, as much as for Google,
or Amazon.
Thus, this difference between us seems more semantic than real. Beyond semantics, however, there is an importantly underspecified part to Frischmannís argument that we should mark, but cannot complete. How should the end-to-end commons be preserved? What norm, or rule, should be imposed to secure it?
Details matter here. Policy debates have bounced between two
very different strategies. One, which was familiar in the late 1990s, is the ěopen
accessî strategy. ěOpen
accessî in this sense was the specific requirement
of network owners that they enable interconnection by competitors to their
physical infrastructure. Thus, open access was the principle under which Digital
Subscriber Line
providers could compete to get access to the wire owned by the telephone company.
The law was held to require the owners of last-mile copper share access to the
customers at the other end.
The aim of this requirement was competition in access providers. Such competition, it was thought, would disrupt the conditions under which it would make sense for network providers to interfere with end-to-end. Open access was thus an indirect means to preserving end-to-end; it was not itself end-to-end.
A second strategy for preserving the Internet commons is more recent. Promoted by the Chairman of the Federal Communications Commission, Michael Powell, this strategy would not insist on physical interconnection.[22] Instead, it would require that network providers secure, as Powell described it, ěInternet freedom.î[23] Four principles defined this ěInternet freedom,î but the essence of the four together is that a network provider not bias or hinder the choice consumers have.[24] Such a regime would assure the network remained ěneutral,î in at least one sense of that word.
Frischmann insists that no network is ěneutral,î which, of
course, is correct. The aim of those pursuing ěnetwork
neutrality,î however, is not some imagined ěneutrality,î
but rather the elimination of certain kinds of discrimination (just as most
policies favoring ěequalityî focus
on rules against certain forms of discrimination). Rules or protocols that have
the purpose of tilting network functionality strategically are the target of
proponents of ěneutrality.î
Scholars such as Tim Wu have outlined relatively simple techniques that might push
help outline such a strategy out.[25]
The choice between these two strategies is hard. The economics of physical interconnection is complex; the inherent games are unavoidable. And more interestingly, it is not even clear that competition in the physical latter would preserve end-to-end neutrality. As van Schewick has demonstrated, under plausible assumptions, competition at the physical layer would not staunch incentives by network providers to bias the network.[26]
Likewise, the mechanics of a ěnetwork neutralityî policy are
not simple either. The neutral network has produced spam and viruses, as well
as instant messagingIM and
Voice over Internet ProtocolOIP.
The network begs for more discrimination, but where, and
how, consistent with ěneutrality,î is impossibly
hard to specify. More importantly, any rule that aims to specify such
neutrality needs to avoid persistent intervention in market operations. The
rule needs to be clear ex ante, so ex post enforcement is feasible.
It is not my aim here to resolve this choice between open
access as a means to the end-to-end infrastructure that Frischmann and I
endorse, and ěnetwork neutralityî as
the means to securing an end-to-end infrastructure. My aim instead is to insist
that there is more to be done to resolve it. Even after we recognize the kind
of infrastructures that needs support, and even if we were convinced that a
commons was the best way to support thoseat
infrastructures, we would be left with the very
difficult question of how we actually construct the commons. That critical
question remains unanswered.
V. CONCLUSION
Simple ideas fix public policy, but simple ideas also break it. A simple idea about how markets in networks should function has fixed U.S. policy regulating those networks. That idea is also slowly breaking those networks. The Internet is not cable TV. The opportunity costs in allowing it to become cable TV are huge. Yet increasingly, government policy is relaxing any regulation that might secure this infrastructure commons. The market, alone, is thought to be a sufficient regulator.
Brett Frischmann has cleared the way
to answering that confusion. His conclusions need to be echoed in the work of
economists; the model needs to be extended to resolve the questions it does not
yet answer. Frischmann has made a critical contribution to an
extraordinarily important debate. We can now see a way to prove what to our
forefathers seemed intuitive—at least until a generation of simple ideas
confused that intuition.